Researchers Link ICD-10 Shift to Financial Losses

The transition from ICD-9 to ICD-10 coding led to significant information loss at one oncology clinic, says a researcher who found that 39 ICD-9-CM codes with information loss accounted for 2.9% of total Medicaid reimbursements and 5.3% of the organization's billing charges.

Healthcare providers will see clinical and billing information and financial losses during the mandated switch to ICD-10 disease classification set later this year, a new study suggests.

The University of Illinois at Chicago study, published this month in the Journal of Oncology Practice, looked at entry ambiguities for hematology-oncology diagnoses in anticipation of the challenges providers may face during the transition from ICD-9-CM to ICD-10-CM, which takes effect on Oct. 1. UIC researchers focused on hematology-oncology because it has fewer ICD-10 codes and less convoluted mappings when compared with other sub-specialties.

The study used 2010 Illinois Medicaid data to identify ICD-9-CM outpatient codes and the associated reimbursements used by hematology-oncology physicians. Researchers identified 120 codes with the highest reimbursement for analysis. They also looked at ICD-9-CM outpatient diagnosis codes and associated billing charges used by University of Illinois Cancer Center physicians from 2010 to 2012 and selected the 100 most-used codes, the study said.

Using a web-based conversion tool developed at UIC, the ICD-9 codes were entered and translated into ICD-10 codes. Researchers looked at whether the translation made sense, whether a loss of clinical information occurred, and whether a loss of information had financial implications.

ICD-10: Minimizing the Financial Hit

"What we found was the transition from ICD-9 to ICD-10 led to significant information loss, affecting about 8% of the Medicaid codes and 1% of the codes in our cancer clinic," said Neeta Venepalli, MD, UIC assistant professor of hematology/oncology and lead author of the study.

Researchers found that 39 ICD-9-CM codes with information loss accounted for 2.9% of total Medicaid reimbursements and 5.3% of UI Cancer Center billing charges.

Venepalli and study co-author Andrew Boyd, an assistant professor in biomedical and health information sciences at UIC, spoke with HealthLeaders Media about the findings and what providers can do to prepare for Oct. 1. The following is an edited transcript.

HLM: Where are these losses coming from?

Boyd: We have three categories: An incorrect mapping, which is flat out wrong, too specific, where you get a whole lot more information, and too general. If you have more detail you can detect fraud. If it is too specific and a patient gets reassigned multiple codes by different clinicians on the oncology side, that could be picked up as fraud by the insurance companies.

You have more specificity, but there is disagreement about that and you get flagged as fraud. The goal is to be revenue neutral from the insurance companies, but with the increased specificity there are other concerns.

This 2.9% (reimbursements) and 5.3% (billing charges) are what you really have to be careful about. This information loss is critical because as a clinician you put down what is medically necessary or correct. But if it is an incorrect mapping and the insurance company maps it incorrectly to 10 and their algorithm to approve what gets reimbursed [may not get you] paid.

HLM: If you are identifying these issues with oncology, which you say is a relatively simple code set, what does this say about potential problems on Oct. 1 with more complex subspecialties?

Boyd: I'm not comfortable making predictions. Right now, appropriately, most of the training for physicians and coders is 'What are the 10 codes I need to memorize?' So, when you're looking at transitioning to ICD-10 the first thing you do is train everyone for the new codes. The first pass of training, just so you can collect money, is training on what the new codes are and what the new interface is.

As ICD-10 Deadline Looms, Providers Fret

What we are talking about in this translation tool and this new paper is that second analysis of these reports that the hospital or the outpatient clinic runs. Right now we are just trying to get through Door No. 1 before we get through Door No. 2. When you begin that second step we're saying here are some problems.

Venepalli: Come October every single report that your hospital is going to generate is going to be under a different set of codes. What are they going to do when they have to compare how that hospital is doing to last year or the year before? They are comparing apples and oranges in some situations.

How do you know? How can you explain your numbers going down or up? How can you use your numbers to predict who you hire or how you should be expanding? This is really relevant for what is happening with (accountable care organizations) and how to you base purchasing. That secondary level of analyses once you are using 1CD-10 may even become more important as a tool to look back at, at least for the first five or six years.

HLM: How can providers best prepare for the transition?

Venepalli: Take your 100 most-frequent billing codes, inpatient and outpatient, and also look at the hundreds you are getting the most reimbursements for and run it through this analysis. What you will find is that the majority of the codes you are OK with and that 18% to 20% of codes are convoluted and maybe incorrect.

This is so easy to do you can do it in an afternoon. Run these codes and wherever you are seeing that the ICD-10 codes are not making sense, or there is some sort of information loss, train your coders, train your billers and physicians to recognize and anticipate that.

Boyd: We have a limited time before the transition and every clinic has a different amount of time to invest in this. If you have 100 codes and it is somewhere between 15 and 20 of codes that are convoluted and you only have a few hours to spend with your staff those are the one you focus on. We are trying to triage the training. You can't spend 100 hours between now and October to train all of the physicians and staff.

The other idea is to have one of your data analysts pull the codes you use for your weekly and monthly reports and if you are actually running your reports off of ICD-9 codes pull those out of the reports and see which reports aren't going to make sense with ICD-10.

If they don't make sense you either have to redesign the report in ICD-10 or just realize that this is complex and some numbers are better than no numbers but this may be incorrect. Physicians and managers are used to uncertainty. Everyone knows you don't know the exact number of patients you are going to see next week.

We are providing tools to help them quantify what the [answers are] in reports and in financials and along those lines. Knowing that 20% of your reimbursement is going to be complex, maybe that is comfortable for you. Everyone has different risk tolerance. Maybe someone is comfortable with that. If not, then spend the hours and the staff time to drill down.

HLM: How does your translation tool work?

Boyd: We built that but it is derived from the government to help with the transition through the General Equivalent Maps. They have the files where they map from ICD-9 to ICD-10 in one file and in the second file from 10 back to 9. We did analytics about future implications. Like everyone else found out, it's hard.

From that we decided to continue to iterate along those lines the way to look at 9 and 10 codes in their totality in both directions so you could understand what the analytical impact was that is how we developed the analysis tool.

Previous guidance for 9 and 10 from the AMA and other agencies told people to only go in one direction, forward or backwards. When we followed their guidance we got conflicting data. The reports changed, which was the impetus to begin this tool. We weren't looking to analyze the mappings for all of them. We took the highest costs and the most complex ones. There is in GEMs something like 150,000 relationships. This is a small project.

Venepalli: When you type in a 9 and try to get a 10, what you get from the current GEMs government-provided mapping is a table of numbers. You don't know what to do with those numbers. I had no idea. It is very confusing. Then this tool represents graphically what these conversions are going to look like. It makes it much easier to look at. We would not have been able to do this analysis to quantify how much of our diagnoses are at risk from information loss and at financial loss only using that table of numbers.

Boyd: We gave the tool away for free and the codes we used to design the tool. If someone wants to import it into an (electronic health record) or if someone wants to make a copy of the tool, we are giving everything away for free. It is not even copyrighted. Please use it.

What's Your ICD-10 HR Strategy?

If someone wants to take the code and bring it in-house because they don't want to post their codes on a website, we have given away in a prior paper the Excel file showing the motifs of every ICD-9 code and the database we used to actually derive the concept of convulsion. If someone has added additional mapping to GEMs they can take their own proprietary mappings and use this algorithm to say we are more convoluted or less convoluted than the government.

HLM: Why is the transition proving to be so difficult?

Boyd: Remember, we've got 500 EHR vendors, several hundred insurance companies, and ICD-9 is used for the medical necessity of service. So, most clinicians will just put down an ICD-10 code saying this is the medical diagnosis.

But if the information is wrong and your insurance company uses the incorrect mapping, we don't know what the insurance company is going to do. You may be able to defend the claim. We are just saying here are the hard codes, here is additional information. Make sure your coders are aware of it. Not everyone is going to be able to go through these 150,000 mappings and make sure everyone is clinically correct or even medically necessary.